CN107241914A - The system and method rewritten for search inquiry - Google Patents

Abstract

Each embodiment includes the system and method for being used to generate query rewrite record, and the query rewrite record can be used for generation to be rewritten for the standardized query of search engine.Such record can identify the constraint and other metadata tokens for rewriteeing and triggering and may be recognized with query rewrite that the rewriting of some of (QRIL) record is associated.In certain embodiments, such record can use that other QRIL are recorded or rewrite information is analyzed, and conflict and generate standardization to prevent from rewriteeing and rewrite.Then the response that generation is inquired about user can be used the information to by search engine.

Description

The system and method rewritten for search inquiry

The reference of related application

The power of the U.S. Patent application No.14/548,105 submitted this application claims on November 19th, 2014 priority Entire contents, are incorporated herein by benefit by quoting.

Technical field

Present application relates generally to the technical field of electronic search, and it specifically related to can be used as a part for electronic search and look into Ask rewriting system and process.

Background technology

In the on-line system that search result is provided is inquired about based on user, often assess what is searched under various factors Object, to produce the search result for the demand for meeting user's request and on-line system.Query rewrite is this search engine On one side.Query rewrite is used to adjust in search the item that uses to match available search result, and in some systems, Query rewrite is mainly responsible for setting up the results set retrieved in response to the search inquiry of user.System described herein and side Method is related to improved query rewrite.

Brief description of the drawings

It is illustrated with the accompanying drawings and nonrestrictive mode shows some embodiments, in the accompanying drawings：

Fig. 1 is the example embodiment of the system for generating, handling and recorded using query rewrite input language (QRIL) Block diagram.

Fig. 2 is shown according to one embodiment for generating, handling and using the details of the QRIL exemplary methods recorded Flow chart.

Fig. 3 is to show the flow chart for being used to generate QRIL exemplary method according to an example embodiment.

Fig. 4 is to show the block diagram according to the generation QRIL of the example embodiment each side recorded.

Fig. 5 shows an example implementation of the QRIL records that can be used together with various embodiments.

Fig. 6 shows one of the search engine that standardized query can be used to rewrite according to embodiment as described herein Example embodiment.

Fig. 7 be show can with embodiment described herein the block diagram for the network delivery system being used together.

Fig. 8 is the example machine that can be used in various embodiments.

Embodiment

Describe the exemplary method and system for electronic search, including the query rewrite system being used together with electronic search The example embodiment of system and process.

Query rewrite is the one side of some search engines.Query rewrite refers to procedure below：It will be received from user The query term synonym relevant with query term or other Given informations matched, and offer be provided be better than By the search result set for the search result that can be provided to the query term application standard search algorithms received.Therefore, inquire about Rewriteeing can work in the search inquiry of processing user and in generation search result set, wherein the search result set Close the results set for being in response to be sent to user in the inquiry of user.

Certain embodiments described herein, which is combined, is used to providing the rewriting system and method for improved query rewrite using looking into Ask and rewrite input language (QRIL) to realize improved query rewrite.For example, search engine can be including individually or group by group The rewritten instructions collection of ad hoc (the ad hoc) of generation, but not used for the system and side for considering influence of the new rewriting to system Method.As additional rewriting is added to this ad hoc rewriting system, the conflict between different rewritings there may be, and be System operator does not know the conflict.For example, when search terms or the associated triggering rewriting value different from two, such conflict It there may be.This may inquire about in response to user according to the mode of application rewriting value and produce unexpected and undesirable search and tie Fruit is gathered.Embodiment described herein the QRIL records that each rewriting can be transcoded into the individual rewriting feature of mark.Then can be with The QRIL is recorded by QRIL processors and handled together with the every other QRIL records in system, collection is rewritten to generate standardization Close.When handling QRIL records, standardization is set up using any overlapping or conflict QRIL records and associated rewriting Structural relation.For example, by QRIL processors according to priority (precedence) rule solve two with overlapping constraint Individual QRIL records, the overlapping constraint instruction should pass through two kinds of different modes rewritten query marks (token).Priority rule Can be based on overwrite type, input time, input entity or any other metadata or mark in QRIL records.So The standardization generated by QRIL processors can be rewritten afterwards and be supplied to search engine for response user search queries.This can To be the same search engine that ad hoc query rewrite set is obtained from it, or this can be different search engines.

Many embodiment search engines have strict service regulation, and service regulation requirement is searched to user in a short time Rope inquiry sends response.Therefore, in certain embodiments must be rather according to the query rewrite system of some example embodiments One of provide rewriting in second or even part millisecond.Due to the time with this correlation of call, this service request is not permitted Perhaps search engine or rewriting system call QRIL database of records or standardization to rewrite database.On the contrary, in some exemplary realities Apply in example, standardized recorded is integrated in search engine system to provide enough query rewrites when receiving user's inquiry Response time.

It is as described herein, rewrite or query rewrite refers to that all or part for inquiring about user used by search engine changes Become or be transformed to another form of translation.Query rewrite at least includes triggering and rewriting value, and the triggering is the value to be transformed Or value set and logical operator, the rewriting value is the transformed value applied to triggering.Query rewrite refers to be drawn by search Hold up using and there is the conversion for being searched the form that engine is used.Its may include using specific file format, text configuration with And the element set of the simplification (streamlined) different from the element set in associated QRIL records.Therefore, this paper institutes The query rewrite used is different from query rewrite input language (QRIL) record, although below putting query rewrite and QRIL records It is discussed in detail together.It is the rewriting for recording establishment according to QRIL by QRIL processors that standardized query, which is rewritten, to eliminate punching Regular collection that is prominent and applying standardization to the rewriting application of QRIL record descriptions.

QRIL records described herein refer to the data structure specific to domain, and it describes query rewrite and on inquiry weight The other information write, the other information enables the system to solve the conflict between different query rewrites, and solve with it is inabundant According to the relevant ambiguity of the query rewrite of the expectation definition of search engine.It is described below in detail for generating and being recorded using QRIL And its system and method for associated query rewrite.

As described herein, user's inquiry refers to the expression user received by search engine system from client device to information Search information.In various embodiments, user's inquiry can take various forms.In a specific embodiment, user Inquiry includes character string.Character string can include multiple words, symbol, space or the numeral of any form.

Search engine mentioned by this paper be configured as receive user inquiry and be available for search engine be used for create with One or more equipment of the search information of the list of the matching of information correlation in user's inquiry.Implemented according to described herein Example, search engine can use any amount of Different matching algorithm.The query rewrite be described in detail herein specifically causes can be by The matching that system operator adjustment is generated by search engine.When the system that can adjust matching algorithm used in search engine is grasped When author carries out similar adjustment, query rewrite enables system operator to carry out such adjustment without making matching algorithm Integrality risk.Weight in adjustment matching algorithm, which has, influences the great wind of matching result in unexpected mode In the case of danger, the standardization for using this standardization described herein to rewrite is rewritten and the embodiment of search engine causes user Matching algorithm can be used to keep intact and keep constant query rewrite to be defined and predictable mode influences to pass through The search result set of search engine output.On the contrary, standardization rewrites the input by matching algorithm is adjusted, to be grasped according to system Self-defined or adjustment search engine operates the ability for make it that its has generation query rewrite the need for author or other systems user. Similarly, be related to data mining, third party, ecommerce sales page, with a large amount of ecommerce sales pages and product phase In the search engine of association and the large-scale complicated system of other system complexity, standardized query rewriting system causes can be by Ecommerce search element spins off from complex search automotive engine system.This is also by making such different each side easily manage Solve the mode of each query rewrite to establish (formalize) query rewrite, and also establish in ad hoc inquiry rewriting and system Other each query rewrites interaction.

Although unstructured to look into addition, the use of query rewrite maintains the integrality of search engine matching algorithm Asking rewriting may collide with each other.For example, the first query rewrite may be by " smart mobile phone A " translates into " product B ".Second looks into Asking rewriting may be by " smart mobile phone A " translates into " product feature C ".3rd query rewrite may be by each note of " smart mobile phone " Number example is translated as " device A ".These query rewrites may be interacted in complicated and uncertain mode.If first Source of the source of query rewrite different from the 3rd query rewrite so that the founder of the first query rewrite does not know that other are overlapping Or the query rewrite of conflict, above mentioned problem especially occurs.Large-scale search engine embodiment can include millions of inquiry weights Write.For example, the search engine of electronic business system is potentially included more than 25,000,000 rewritings.Embodiment described herein offer Standardizing priority order rule, the standardizing priority order rule determines how the difference for solving to be operated in same system and looked into Ask the conflict and interaction between rewriteeing.

Embodiment described herein each side be related to the classification of overwrite type.Specific embodiment can use overwrite type Different classifications.As discussed herein, " direct " or basic rewriting are a kind of overwrite types, and it is a kind of rewriting class that phrase, which is rewritten, Type, " mark refinement " is a kind of overwrite type, and " full query rewrite " is a kind of overwrite type.Other realizations can include weight Other classification write.

It is as described herein, it is directly overwritten the triggering for being related to and being directly associated with rewriting value.Although being directly overwritten to have In terms of other associations (including various constraints, classification and metadata), but basic structure is the direct pass between triggering and rewriting value Connection.The example for being directly overwritten structure including the additional information associated with E-commerce Search Engine is：

Constraint=[send the e-commerce website of inquiry, trigger, classification constraint, inquiry starting country]

Rewrite=[rewriting value, classification is rewritten, and aspect is rewritten, bulleted list position ID]

It is using the rewriting example of said structure：

Constraint=[send e-commerce website=" U.S. " of inquiry, keyword triggering=" fone ", classification=" electricity Son ", inquiry starting country=" Canada "]

Rewrite=[keyword is rewritten=" smart mobile phone ", and classification rewrites=" 123456 ", aspect rewriting="None", project List placement ID=" Canada "]

As it is used herein, phrase rewrites to be related to is rewritten as phrase by triggering, wherein, phrase is defined as continuous word The sequence of mark.This is with the difference being directly overwritten：The rewriting value that may have as single mark is directly overwritten, and The rewriting value that phrase is rewritten is the phrase for being related to multiple marks.In addition, although the weight with multiple marks may be had by being directly overwritten Phrase is write, but the mark of the not specified rewriting value being directly overwritten order.Compared with corresponding be directly overwritten, phrase, which is rewritten, to be caused More specifically project set can be called.For example：With with triggering " built-in camera " and rewriting value " built-in PHRASE is (rearmounted Camera) " phrase rewrite and compare, the direct weight of " built-in camera " and rewriting value " built-in rearmounted camera " is triggered with identical More projects will be matched by writing.In some cases, it is preferably as it will match more accurate Item Sets that phrase, which is rewritten, Close.

As it is used herein, mark refinement refers to add or be removed to triggering the rewriting of keyword.For example, If triggering " cheap new princess's smart phone shell " is without acceptable search result set is provided, system can make Rewritten with mark refinement and word (that is, mark) is deleted from inquiry.If sufficiently frequently seeing that this touches in user inquires about Hair, then system, which can be collected, is enough to determine that mark " cheap " and " new " attempt retrieval usually using the search inquiry for user Search result element for the information that is not critical.Therefore, deleting that these mark refinement can be used for will be " cheap new Princess's smart mobile phone shell " is rewritten as " princess's smart mobile phone shell ".This is the example that mark is abandoned.On the contrary, mark is refined It can be used for adding word to inquiry.For example, " brand A " may have mark refinement rewriting value, and " model #123 ", this can for triggering It can be unique welcome product associated with brand A in E-commerce Search Engine.Mark refinement is the change to triggering, Rather than the tradition rewriting of the trigger value in user's inquiry is replaced using rewriting value.Although some mark refinements in some cases May have and be directly overwritten identical functional effect, but in various embodiments, create thin with mark for being directly overwritten Conflict solving and certain form of QRIL interrecord structures can be realized by changing the classification rewritten.

Fig. 1 is to be used to generate, handle and using the example of the system 100 of query rewrite input language (QRIL) record implement The block diagram of example.QRIL is specific for the language in domain.Specific embodiments described herein can provide QRIL, and it is to be used for electronics business Business search and specific to ecommerce search inquiry rewrite the language specific to domain.Although described herein some exemplary Embodiment and ecommerce and ecommerce search inquiry are especially relevant, it will be obvious that embodiment described herein it is each Aspect can apply to other kinds of search inquiry domain.

System 100 includes query rewrite source 110, inquiry transcoding device 120, QRIL database of records 130, QRIL processors 140th, product database 150 and search engine 160.Standardization rewriting set 142, which is also shown in, is transmitted to search engine 160th, the output of product database 150 or the QRIL processors 140 of both.

As shown in system 100, query rewrite source 110 includes multiple different rewriting sources.It may include any amount of institute The example shown rewrites source and other kinds of rewriting source.Query rewrite source 110 is shown as including inquiry database 112, number According to excavation module 114, rewrite optimization module 116 and editor's web service module 118.

Inquiring about database 112 includes local data base, and the local data base has the spy from various networking data storehouses source If the set of query rewrite or ad hoc query rewrite.For example, inquiry database 112 can include query rewrite set or with search The different search engine of engine 160.This can include the different query rewrite forms and/or knot used from search engine 160 The search engine of structure.The information can be sent to inquiry transcoding device 120 as query rewrite data acquisition system.

The system that data-mining module 114 includes analysis the following：User inquiry, as search engine to these users The search result of the response of inquiry and user's selection after user receives search result.Such user selection can be with Including：The link of specific website, the product that user's purchase is listed in search result are chosen, or receives and uses in user The user action for any other record taken after the associated search result of family inquiry.Such data can be wrapped additionally Include the information on the different query rewrites from submitting the different user of identical initial search query to be used together.Using sufficiently large Data acquisition system, can for specific input inquiry, query rewrite, search result and user response generation statistical information and point Analysis.Data-mining module 114 can analyze such information to generate query rewrite data acquisition system.

Rewriteeing optimization module 116 includes the rewriting database of such as product database 150 etc.For example, can be to rewriting Optimization module 116 transmits the standardization from product database 150 and rewrites 142.Rewrite optimization module 116 and then mark can be analyzed Standardization is rewritten 142 set to recognize the rewriting of poorly efficient redundancy, or deposited based on the part that set 142 is rewritten as standardization Rewriting generate new rewriting.The new rewriting that is recognized by optimization module 116 or by rewriting optimization module 116 recognize it is any Redundancy or poorly efficient rewriting can send inquiry transcoding device 120 to as query rewrite data acquisition system.

Editing Web service module 118 includes service door, and the service door allows third party to access system 100 with life Rewritten into the QRIL records of customization and associated standardization.For example, editor's Web service module 118 can include registration service Device, the registrar enable on the electronic commerce gate associated with search engine 162 sell product trade company Query rewrite data acquisition system is submitted to inquiry transcoding device 120.In such embodiments, trade company can be related to specific constraint Connection.For example, trade company can have the StoreFront OR gate of the part as the e-commerce site associated with search engine 160 Family.The QRIL records of the query rewrite data acquisition system generation provided according to trade company can automatically include according to those QRIL being remembered The standardization of record generation rewrites the constraint for the StoreFront for being restricted to trade company.Further, since QRLL processors 140 realize that priority is advised Then, system 100 is limited probably due to coming from third-party query rewrite data via what editor's Web service module 118 was received The latent fault gathered and introduced.

Inquiry transcoding device 120 receives the query rewrite data acquisition system from query rewrite source 110, and is given birth to using the information Into QRIL records.Such QRIL records specially can rewrite the information generation that source 110 is received, Huo Zheke according to from single query With according to the query rewrite data generation QRIL records received from multiple sources.Craig is come from some embodiments it is possible to combine The historical data that the query rewrite data in rewriting source 110 are stored using inquiry transcoding device 120, to generate QRIL records.Below Inquiry trans-coding system 400 on Fig. 4 discusses the additional detail related with inquiry transcoding device 120 to inquiry transcoding.

Once generating one or more QRIL by inquiry transcoding device 120 to record, then QRIL records are stored in QRIL notes Record at database 130.QRIL database of records 130 can be with inquire about transcoding device 120, QRIL processors 140 or it is any its The memory storage device of his integration of equipments.The storage of QRIL database of records 130 can be used for generation standardization to rewrite set (example Such as, standardization rewrite 142 set) QRIL set of records ends.In certain embodiments, QRIL database of records 130 can be wrapped Include different QRIL set of records ends.This can make single query transcoding device 120, QRIL database of records 130 and QRIL processing Device 140 can provide different standardization for different search engines and rewrite set.

When search engine 160 to be directed to generation standardization, which rewrites 142, gathers, QRIL processors 143, which are used, comes from QRIL The QRIL records of database of record 130.In some embodiments it is possible to each QRL records are retrieved respectively, or can be by QRIL Processor 140 disposably asks QRIL set of records ends.Then, the analysis of QRIL processors 140 comes from QRIL database of records 130 QRIL set of records ends, with generate standardization rewrite 142 gather.As a part for the process, it can recognize and each QRIL The associated overwrite type of element, and can handle other constraints and/or meta-tag information rewritten with not only to generate standardization but also Solve any conflict between the standardization rewriting of different Q RIL element definitions.It is QRL processors that standardization, which rewrites 142 set, 140 output, it is analyzed the QRIL records from QRIL database of records 130 by QRIL processors 140 and produced.Work as standardization When rewriteeing the completion of set 142, it can be exported to product database 150 from QRIL processors 140.In various embodiments, produce Product database 150 is optional.As described above, product database 150 can be used for the reality that checking is activated in search engine 160 Standardization is rewritten.Product database 150 can also be rewritten optimization module 116 and be used for rewriteeing 142 set later in the standardization Further refinement is rewritten in more redaction.Product database 150 can also be used together with test search engine, some to verify QRIL records are to standardization rewriting and the influence of the search result associated with rewriteeing rewritten user's inquiry by standardization. For example, editor web service module 118 can provide to test search engine (not shown) trade company access and pair with product number The rewriting of 142 set is rewritten according to the related standardization of the trade company in storehouse 150.Editor's Web service module 118 can enable trade company Will be by Corey transcoding devices 120 and QRIL processors to provide based on trade company's change from trade company's query rewrite data acquisition system The query rewrite data acquisition system of 144 processing, i.e., nonproductive (nonproduction) standardization rewrites set.Then, trade company can be with Submit test query and rewrite how the standardization with previously existing rewrites interaction to observe these test queries, with test Generation search result set in search engine.

Search engine 160 can be that any search for the query rewrite for rewriteeing 142 set etc using such as standardization is drawn Hold up.Draw as described above, specific ecommerce relevant search is described in detail (particularly in Fig. 8 search engine 800) herein Hold up.In a specific embodiment, search engine 160 is E-commerce Search Engine, and it is used as search matching algorithm The product category tree of a part comes according to user's query generation search result.Because related to such E-commerce Search Engine The electronic commerce gate of connection is directed into the product that identification is bought for user, this species being integrated together with search engine 160 Be able to Shu not provide can be integrated with query rewrite constraint and meta-tag information, to build product in E-commerce Search Engine Search and search result.For example, such classification tree can make the brand name associated with some product types match pin Search to product type.This matching may not be carried out in the system using text or word matched algorithm.Below more It is discussed in detail with this E-commerce Search Engine about and with standardized query rewriting and associated QRIL records The information-related additional detail of classification meta-tag.

System 100, which is described, to be rewritten for the standard for generating QRIL records and being associated and uses basis in a search engine One of the system that the standard of QRIL record generations is rewritten may realize.In various embodiments, each element of system 100 can With the module being implemented as in individual equipment or multiple equipment.Such element can also be implemented as separation equipment or across The system of multiple equipment operation.Therefore, inquiry transcoding device can be with operating in the identical equipment of QRIL processors 140 Module.Or, inquiry transcoding device 120 can be the network system of computing device, and it is with constituting the one of QRIL processors 140 Individual or multiple equipment network apart from each other.

Fig. 2 is shown according to one embodiment for generating, handling and using the thin of the QRIL exemplary methods 200 recorded The flow chart of section.Although such method 200 can be performed by the various not be the same as Examples of innovation presented herein, in order to The purpose of explanation, the operation of method 200 is described in the context of system 100.

Operation 205 is the above-mentioned optional registration step on editing Web service module 118.Such registration can make certain A little system users can generate the QRIL records with binding occurrence, and the rewriting associated with QRIL records is restricted to by the binding occurrence Particularly relevant search rather than generation QRIL records with system user.The example of such association can be the virtual shop of operation The trade company in face, it has the access right to the wider delivery system of such as system 700.Such QRIL records can include Associated standardization is rewritten to the constraint for being limited to be only applied to the inquiry from the virtual storefront reception of trade company.When to third party (such as trade company, search consultant, system user, middleware provider or any other such third party) is provided to system 100 Access right when, operation 205 can be carried out.Operation 205 is the registration to inquiry transcoding device 120.In various other embodiments In, middle editor's Web service module 118 can fully handle Accreditation System, or can present for handling registration, visiting Ask the added layer of security and user interface layer with other various account details.In other embodiments, the He of query rewrite resource 110 Inquiring about transcoding device 120 can be communicatively coupled as a part for network or other some communication paths, without correlation The registration process of connection.

In operation 210, query rewrite data are received by inquiry transcoding device 120.This can be in response to select in operator Select, or as system update a part be periodically supplied to inquire about transcoding device 120 query rewrite data from It is dynamic to update.Query rewrite data are supplied in the embodiment of transcoding device 120 in response to operator's selection, can be by operating The user of the machines such as third-party server 730, client machine 710 or client machine 712 (enters below to make a choice Row more detailed description).It is used as a part for the operation of such equipment, third party application 728, web client 706 or journey Sequence client 708 can be included with the user interface for inputting selection, and it allows users to send out to inquiry transcoding device 120 Send query rewrite data.Such application or client can be entered with inquiry transcoding device 120 or middle registration equipment or application Row communication, is registered using the part as prior operation 205 to system 100.Automated communications for query rewrite data or The option of user's selection communication of query rewrite can be selected as a part for registration by user, or can be by predetermined System set set automatically.

In operation 215, the query rewrite data received in operation 210 are analyzed, to recognize what is triggered and be associated Rewriting value.As it is used herein, triggering refers to character, word, phrase, symbol or any other information aggregate, triggering as When the part inquired about for user is received, these information aggregates are converted using the part as rewriting for initiating to rewrite Into another form.For example, " smart mobile phone " one word can be triggering.And associated rewriting value is probably " A participants in a bridge game's machine ".Such as Fruit receives the query rewrite data including query rewrite set from inquiry database 112, then can be in query rewrite data Explicitly mark is triggered and associated rewriting value.In this case, character resolution device can be used for from query rewrite data Identification triggering and rewriting value.If the character resolution device for analyzing query rewrite data does not recognize triggering explicitly and associated Rewriting value, then can perform analyzing adjuncts with recognize rewrite in triggering, or inquiry transcoding device 120 can by inquiry weight It is the data not comprising triggering or rewriting to write data markers, and QRIL records can not be created according to the data.Below with reference to Fig. 4 description of inquiry transcoding device 400 and triggering recognizes the additional detail relevant with identification is rewritten.

In operation 220, query rewrite type is distributed to the triggering and rewriting identified.Query rewrite type is used for It is determined that standardization rewrites the priority or priority grade that will be received derived from query rewrite data.Query rewrite type Determined by the structure rewritten, support data or with the metadata of the associated part as query rewrite data of rewriting, or Thus the two is determined." smart mobile phone " above-mentioned rewriting associated with rewriteeing " brand A phones " is referred to herein as directly Rewrite.The structure being directly overwritten includes triggering and rewriting value.This is simplest structure, is replaced wherein rewriteeing and being related to rewriting value Triggering.Other examples of overwrite type include phrase rewriting, mark refinement and rewritten and full query rewrite.On query rewrite type It will be discussed with the additional detail of the specific example embodiment of different query rewrite types below with reference to Fig. 4.

In operation 225, any other related constraint being present in query rewrite data or metadata letter can be recognized Breath.Identification with the triggering in rewriting value is similar, and other related constraints or metadata information can exist as set character In data, and character resolution device can recognize the word that the known some constraints relevant with the elements that record of QRIL match Symbol packet.Each composition that QRIL shown in Fig. 7 records 700 and QRIL records 700 is shown including multiple different QRIL The example QRIL records of element.As a part for operation 220, resolver can recognize any composition of QRIL records 700.

In addition, in operation 225, according to the triggering identified in operation 215 and 220, rewriting value, query rewrite type And related constraint or metadata information generation QRIL records.Such record can use processor to generate, to create note Directory structures and resolver collection text, symbol or other operator messages used from operation 215 and 220.Below will It can be used for the additional detail of the system of QRIL record generations with reference to Fig. 4 discussion.Certain amount of member can be used usually to construct QRIL is recorded, or the element with the related information from query rewrite Data Identification can be used only to generate QRIL records. If QRIL records are usually constructed using the member of set number, and not for the first usually identification letters of specific QRIL Breath, then can use the member with zero or empty entry usually to construct QRIL records.As described above, QRIL 700 is QRIL records Example, and in one implementation, 700 can be recorded by generation QRIL during operation 225.Be discussed below has with QRIL records The additional detail of pass, is discussed especially with regard to Fig. 5 QRIL records 500.

In operation 230, whether systems inspection can generate additional QRIL according to the query rewrite data received is remembered Record, or rewrite data with the presence or absence of the additional queries to receive and analyze from one or more query rewrite sources 110.If also Additional query rewrite data are analyzed, then process continues in operation 235, handle the additional rewriting from one or more sources Data.The query rewrite data can come from single query and rewrite source 110 or multiple queries rewriting source 110.It can include making For any source shown in the part in query rewrite source 110, the query rewrite source 110 includes inquiry database 112, data mining Module 114, rewriting optimization module 116 or editor's Web service module 118.Then 215 to 230 are repeated, can be concluded that does not have The additional query rewrite data of residue will be handled.In various embodiments, any amount of processor, inquiry transcoding can be used Equipment 120 or the next this query rewrite number performed simultaneously in operation 210 to 230 of other modules or device for performing this operation According to processing.In other embodiments, the processing can be the operational set periodically performed, or whenever receiving identification The operational set performed during the triggering of new query rewrite data.In some embodiments it is possible to raw in operation 210 to 230 Record and polymerize into QRIL so that the QRIL records generated in different time are all sent to QRIL record datas together Storehouse.In other embodiments, each QRIL records are stored in when it is generated in QRIL database of records 130.Some In embodiment, single query transcoding device 120 can send QRIL records to multiple databases, and can be by being remembered according to inquiry The information that record data are identified, the mark by query rewrite source 110, or the part registered in as operation 205 are received To information determine target database.

If not identifying additional rewriting data in operation 230, all QRIL are recorded in operation 240 It is stored in QRIL database of records 130.The QRIL records being stored in QRIL database of records 130 can be stored for Use later so that existence time postpones between operation 240 and operation 245, or is stored in QRIL databases 130 more New and new QRIL records can be immediately transmitted to QRIL processors for analysis.

In operation 245, by one or more QRIL processors analysis QRIL records.In some embodiments it is possible to by Single QRIL processors serial analysis each QRIL records.In other embodiments, can be by one or more QRIL processors (such as QRIL processors 140) parallel parsing QRIL is recorded.The analysis of QRIL processors determines the form associated with search engine, And the information recorded from QRIL, the information is needed for being rewritten with the acceptable form generation standardization of search engine.Though Right method 200 describes the example implementation that the analysis of QRIL processors and standardized query are generated, but below with reference to Fig. 5's QRIL processors 540 description can in different embodiments using QRIL processors operate additional detail and other for the use of.

As a part for operation 245, it can recognize that what the element recorded as QRIL included looks into by QRIL processors 140 Type is ask, and a variety of processing operations can be realized based on the query type of QRIL records.Be discussed below from it is different The associated details of query type, and during operation 245, QRIL processors 140 can use the spy of different query types Levy.It is as discussed in more detail below after carrying out initial analysis to QRIL records in operation 245, rewriting conflict can be used and preferential Cahn-Ingold-Prelog sequence rule come operation 250 to 290 in generate one or more rewritings.

In addition to different types of query rewrite discussed above, some QRIL records and associated rewriting may be related to And recurrence is rewritten.Operation 250 checks QRIL records to obtain the setting related to recurrence rewriting.Term recurrence rewrite refer to when with First rewrites the rewriting chain that may occur when associated rewriting value is the triggering associated with the second rewriting.If for example, One is directly overwritten with triggering " fone " and rewriting value " smart mobile phone ", and second is directly overwritten with triggering " smart mobile phone " " telephone model #12345 ", then rewriteeing chain may cause the mark " fone " in user's inquiry to be rewritten as " phone with rewriting value Model #12345 ".The inspection of operation 250 can be related to QRIL record elements, its indicator pair weight associated with QRIL records Whether write allows or enables recurrence rewriting.In certain embodiments, some QRIL records can include especially allowing or special Ground forbid rewriting value be used as another rewriting triggering element.In other embodiments, system convention may determine whether to permit Perhaps recurrence is rewritten.If allowing recurrence to rewrite, system may proceed to what is analyzed any related QRL records or be previously generated Standardization is rewritten.If for example, the use of the rewriting of triggering " fone " being to indicate not allow one of the QRIL elements of recurrence rewriting Point, then standardized for the QRIL Element generations rewrite during ignore the of the part that set is rewritten as standardized query Two rewrite.If however, allowing recurrence to rewrite, in operation 255, QRIL processors 140, which will be checked, to be had with handling QRIL elements the rewriting of triggering that matches of all or part of rewriting value.This may include to check QRIL database of records 130 In all QRIL elements.This can also refer to check and is rewritten from the current standardized all standardization for rewriteeing set 142.If Applicable rewriting is found during operation 255, then analysis recursion is rewritten in the repetition of operation 245.Then in operation 250 Check that recurrence rewrites to obtain double recursive rewriting in repeating.The process is carried out in a nesting relation, until no other recurrence Rewrite, or until reaching the system limitation that recurrence is rewritten.In certain embodiments, single QRL can have touches from identical Two of hair are nested to be rewritten.If for example, the first QRIL has rewriting value " Belgium double chocolates ", and for " Belgium " All there is applicable triggering in " double chocolates ", can be to " Belgium " and " double chalks if then prevented without other constraints Power " nested rewrite of the two is analyzed, and by its respective rewriting value in establishment standardization is rewritten.

In act 260, QRIL processors 140 may determine whether to conflict with any of rewriting.The example of conflict Including the rewriting with identical triggering and different rewriting values.Operation 260 can be related to QRIL processors 140 relative to other QRIL Record, rewrites relative to the standardization being previously generated for being incorporated to standardization rewriting set by QRIL processors 140, or relatively In both it come check to current QRIL record rewriting.

If recognizing entry/exit conflicts, in operation 265, the rewriting in network analysis conflict, and application priority rule To solve conflict.It is associated when identical is triggered from multiple different rewritings, and one rewrites and one or more rewriting punchings , it is necessary to solve conflict when prominent.This is likely to occur in for example when phrase rewrites and is directly overwritten with identical triggering, and this is identical tactile Hair has identical mark.In this case, phrase is rewritten only with passing through the corresponding project for being directly overwritten matching generally Subset matches.System can solve this conflict with unalterable rules.One embodiment is defined when phrase is rewritten and directly heavy Write when being triggered including identical, be directly overwritten and be dropped, and phrase is used by system when providing more succinct matching set Rewrite.Another embodiment assesses the expected results set rewritten from two conflicts.Selection has a greater amount of rewriting details Rewrite, a greater amount of rewriting details can be expected to return to narrower search result.Can be based on the character in rewriting value or note Number quantity this is estimated.Meta-tag value or other relevant informations in can also being recorded based on QRIL is come to this progress Assess.For example, QRIL records can include the meta-tag element for priority fraction or detail value.Such meta-tag value Available for parsing when triggering be a part for search inquiry when QRIL record or applications trigger order.

In a potential embodiment, priority is assigned for the query type of some conflicts based on query type. In a potential embodiment, full query rewrite is endowed priority, because full query rewrite is and user's inquiry string Accurate matching.Full query rewrite is rewritten including no derivative transformation (derivative transformation) or recurrence Specific rewriting value, because full query rewrite is specific to accurate user's inquiry and tailors.Therefore, full query rewrite It is preferential rewriting, and in view of the priority of full query rewrite, is not carried out the rewriting of different types of any conflict.Because Full query rewrite has inquires about the triggering accurately matched with user, and the full query rewrite of conflict may result in the mark quilt of mistake Export to Systems Operator.In the embodiment of not this conflict output error, as described above, entirely inquiry conflict can be by Solve, the rewriting value comprising the maximum amount of details obtains priority.

The example embodiment of the Conflict solving of full query rewrite is continued with, in the present embodiment, mark adjustment rewriting can Priority after then full query rewrite is obtained, and phrase rewriting can obtain excellent higher than what is be directly overwritten as described above First sequentially.Any rewriting conflict between the rewriting of same type can be solved as described above, to be conducive to most narrow rewriting value. If allowing derivative (derivative) rewriting or recurrence to be rewritten so that the rewriting value of the inquiry of rewriting can be as additional heavy The triggering write, then each recursive levels after the rewriting completed can use rule same as described above each to solve The rewriting conflict of derivative rank.

When it is determined that during Conflict solving, all related conflicts, which rewrite to rewrite in standardization as a part for operation 275, to collect It is updated in conjunction.In certain embodiments, this can be related to one for rewriteeing and being removed in set in the rewriting from standardization.At it In his embodiment, this is related to selection rewriting order so that will use the first rewriting, and convert triggering with rewriting value Afterwards, another triggering will not be reapplied.In a potential embodiment of priority regular collection, full query rewrite, which has, to be higher than The priority of every other rewriting, is directly overwritten with the priority rewritten higher than phrase and mark is refined, and phrase weight Write with the priority refined higher than mark.Can be based on the level of detail in rewriting value (for example, character, mark or symbol Quantity) carry out rewriting to same type and assign priority, wherein, higher detail grade (for example, more characters) has height In the priority of relatively low amount of detail.In certain embodiments, QRIL can be between the rewriting for being used to solve same type The priority meta-tag element of conflict.

In operation 270, after considering all elements of QRIL records and having solved any conflict, by The generation standardized query of QRIL processors 140 is rewritten.In operation 280, update standardized query and rewrite set with including new Query rewrite.In operation 285, QRIL processors 140 check whether any additional QRIL records will be taken into account and be used for The standardization that generation will act as the part that standardized query rewrites set is rewritten.The processing performed by QRIL processors 140 is right After repeat 245 to 285, until considering all applicable QRIL records.

When terminating to consider all QRIL records, rewrite 142 from the outputting standardization of QRIL processors 140 and gather.Various In embodiment, this can be the output communication of cache from QRIL processors 140 or local storage.In other implementations In example, this can be the final adjustment carried out by 140 pairs of texts being stored in single memory of QRIL processors, its In, text includes standardization and rewrites 142 set.In operation 290, it is supplied to search to draw standardized query rewriting set Hold up 160.In operation 295, search engine rewrites set to operate using standardized query, with response to being connect from client device The inquiry that receives generates search result.Search engine proceeds, until as operation 298 a part of generation system more Newly.When occurring system update, the process can be repeated since operation 230, generate new QRIL records, processing QRIL records The standardized query rewriting set that set and renewal search engine 160 are used is rewritten to update or generate new standardized query.

Fig. 3 is to show the flow chart for being used to generate QRIL exemplary method 300 according to an example embodiment.Method 300 can be performed by a part for bigger system, be performed as the module in computing device, or by inquiry transcoding device (example Such as, inquiry transcoding device 120 or inquiry transcoding device 400) perform, inquiry transcoding device 120 or inquiry transcoding device 400 are respective Including storage device, input and output module and it is coupled to storage device and input and one or more processing of output module Device.

Fig. 4 is coupled to the block diagram of QRIL database of records 130 and the inquiry transcoding device 400 in query rewrite source 110, with On QRIL database of records 130 and query rewrite source 110 both as system 100 be with a portion of into discussion.Exist below Method 300 described in exemplary embodiment using Fig. 4 inquiry transcoding device 400.

Method 300 receives the first inquiry at inquiry transcoding device since operation 305 from the first query rewrite source device Rewrite data acquisition system.In the exemplary embodiment of inquiry transcoding device 400, the query rewrite is received at input module 422 Data.Query rewrite data acquisition system can include any information related to product or search, and including bound data, member mark The evidence that the counts query rewrite data related to any other.Query rewrite data include can be used for the first trigger value of mark and correlation The information of first query rewrite value of connection, the first query rewrite value is constituted together with the first trigger value by the core letter as rewriting Breath.Bound data can be used for the appropriate limitation that mark is rewritten to correlation.Meta-tag data include being worth it with actual triggering and rewriting Outer overwrite type related any information or data.Meta-tag data can also include indicating whether recurrence rewriting is allowed to use The data rewritten in correlation, can assist mark associated with rewriting in the case where classification constraint is not identified explicitly Classification data, or different from other classifications of the possibility that explicitly the identifies classification associated with rewriting.

In operation 310, Query Information is handled to recognize the first triggering and the first query rewrite value.In inquiry transcoding device In 400 example embodiment, data parser module 424 can be used to complete the processing.Data parser module can be Analysis inquiry data provide the text resolver or other calculating resolvers for the data structure that inquiry data are represented to build.Data Character or symbol in resolver analysis inquiry data, to recognize that triggering and rewriting value are used as the core of rewriting, the core Center portion point would is that the basis of QRIL elements.Data parser can also use mark or character repertoire and lead to recognize in inquiry data Cross the associated matching score of other elements or character string that the storehouse to record with some metadata, constraint or QRIL.

Operation 315 then be related to analysis first inquire about data acquisition system with recognize in multiple queries overwrite type with first inquiry Rewrite the first associated query rewrite type of data acquisition system.In one embodiment, generated by data parser module 424 Data structure can be used in combination with multiple overwrite type identifier modules, recognize the query type associated with data query. Rewrite identifier module 428, mark refinement identifier module 431 for example, being directly overwritten identifier module 426, phrase and look into entirely The storehouse mark or structural information relevant with the overwrite type characterized by rewriting system can each be included by asking refinement identifier 431. As data parser module 422 analyzes query rewrite data, module can use next free data parser module 424 to analyze With the information of the query rewrite data of construction, query rewrite data are associated with query type.If parsed using data The module 426-431 of device module 424 does not identify query type, then QRL generations and formatting module 436 can determine not root According to query rewrite data generation QRIL records.

Except recognize in operation 310 the first triggering in the first query rewrite value and in operation 315 identification first look into Ask outside overwrite type, additional embodiment can analyze query rewrite data to obtain other information.These other informations may Including the details for creating meta-tag, for the constraint that system is told when being used or be not used is carried out in rewriting The details of identification, or other this type of informations.Additional detail on this meta-tag is discussed below in reference to Fig. 5.Can be with Using by data parser module 424 identified knot is combined with inquiring about other any amount of modules of transcoding device 400 Structure collects these information.This includes constraint identifier module 432, and it especially can be adapted to and be configured with analysis system, with Recognize the constraint to specifically rewriteeing.The example of this constraint is included only when inquiry is derived from specific country or other geographic areas Client device when just application rewrite, the just application rewriting only when inquiry is derived from specific website market, or any other this The constraint of sample.For example, query rewrite data can be derived from specific query rewrite source (for example, editor's web service module 118).Make For a part for registration process, editor's Web service module 118 can be registered to inquiry transcoding device 400, and this can be to number Mark value is provided according to parser modules 424, the mark value enables data parser module 424 to recognize from editor's Web clothes The rewritten query data for module 118 of being engaged in.When the data parser module 424 by being operated together with constraint identifier module 432 is known When not going out the mark value, for QRIL record generation binding occurrences, associated query rewrite is used for and volume by binding occurrence limitation Collect the associated user's inquiry of Web service module 118.This can include having drawing using editor's Web service module 118 and search Hold up the trade company of the e-commerce website both 160, search engine 160 is from including the system acceptance criteria of inquiry transcoding device 400 Change query rewrite.Similarly, can be generated by the data parser module 422 operated together with labeled identifier module 434 can The metadata associated with the meta-tag element that QRIL is recorded.In addition, each in these modules can use overwrite type One in identifier module 426-430 is probably unique mark for specific overwrite type or constrains to recognize.

Then, operation 320 is related to generates the first query rewrite input language according to the first query rewrite data acquisition system (QRIL) record.Then, operation 325 is related to stores the first QRIL notes in the QRIL database of records recorded with multiple QRIL Record.First QRIL records include the first trigger value and the first query rewrite value.It can be made by QRIL generations and formatting module 436 Recognized with any module using above-mentioned inquiry transcoding device 400 or the value of generation records to generate QRIL.Then, as operation 325 part, output module 437 can send the QRIL records that QRIL is generated and formatting module 436 is generated to QRIL Database of record 130.In certain embodiments, QRIL records can be included except the trigger value of core element first and the first inquiry weight Write the additional elements beyond value.

Fig. 5 shows an example implementation of the QRIL records that can be used together with various embodiments.Fig. 5 QRIL notes Record 500 shows the example of the QRIL records with multiple additional elements.In some embodiments it is possible in the reality of operation 320 QRIL records 500 are generated by QRIL generations and formatting module 436 in existing.QRIL records 500 include triggering element 540 and rewritten It is worth element 550.In addition, QRIL records 500 include multiple constraints 510, the plurality of constraint 510 includes class elements 512, website member Element 514 and country elements 516.Class elements 512 can identify the one or more classes associated with E-commerce Search Engine Not, the query rewrite associated with QRIL 500 will be applied to the E-commerce Search Engine.As being described in detail further below, Based on the details and the other information relevant with user or user client device of these inquiries, search engine can be based on electronics Classification in commercial search engine classification tree is classified to user's inquiry.Category information can additionally be used for inquiry weight Write.The example used is that the classification that identification is identified by class elements 512 is constrained.

Similarly, search engine can have the letter that be derived from national or other geographical position are inquired about on user Cease, and this can be used together with country elements 516 with when inquiry is come from by specific QRIL records (for example, QRIL is recorded 500) to use or enter row constraint without using some query rewrites during the position that country elements 516 are identified.

Site element 514 can identify the website of another constraint that can serve as specific query rewrite, merchant stores or its His electronic commerce gate.For example, in one embodiment, the multiple ecommerce of 720 trustships that system 700 can be applied via market Market.There can be site identifier using 720 associated each markets with market.The site identifier may be used as QRIL The value of site element 512 in record 500.This can enable specific market to create QRIL records using 720 operator 500, and identified using site element 514 to be constrained to QRIL records 500 to be only applied to be derived from by the value of site element 514 Trade company market apply 720 inquiry.

The meta-tag 520 of QRIL records 500 can include the QRIL elements for any number of different types of information. In Fig. 5 QRIL records 500, meta-tag 520 is used for the query rewrite type for identifying the rewriting associated with QRIL 500.This is By providing what member was usually completed for each query rewrite type.Therefore, QRIL records 500 include that full inquiry weight can be indicated The full inquiry element 526 of type is write, can indicate to be directly overwritten the direct element 528 of type, can indicate that mark refines type Mark Refining Elements 534 and can be with the phrase element 536 of referring expression overwrite type.

In addition to query rewrite type, meta-tag 520 also can indicate that with rewriteeing associated other of QRIL records 500 Details.Exclude element 530 and can serve to indicate that some overwrite types are passive rather than positive.This means rewriting be for Exclude the search result comprising rewriting value rather than carried out to search for the result for including the rewriting value.Derivative rewrite is prohibited It can be used for recognizing that recurrence is rewritten with element 532 whether to be allowed to use the rewriting value of rewriting value element 550 and be used as follow-up weight The triggering write.Categorical match 539 and phrase categories 538 can be identified will be with search one in the classification tree of E-commerce Search Engine The classification used is acted, the search is performed using the rewriting value of rewriting value element 550.In other embodiments, it can make It is used as the part of QRIL records (for example, QRIL records 500) with any amount of other elements.

Fig. 6, which is shown, can use the search engine system rewritten according to the standardized query of embodiment as described herein 600 example embodiment.For example, system 600 can receive the search that standardization rewrites set from QRIL processors 140 The realization of each side of engine 160.In certain embodiments, system 600 can be the electronics business with such as Fig. 7 system 700 Business platform or the associated E-commerce Search Engine of delivery system.

Online resource is accessed for user, the commonly provided search service of provider of such as e-commerce website provider is come The positioning resource related to user interest.The target of provider is to provide the result for meeting some concerns of user and provider, User is for example caused to reuse the correlated results of provider there is provided the extra earning of business and meet business partner (for example, advertiser Or sponsor) concern.When provider is ecommerce provider, it is considered to for example paid close attention to from search result or industry partners Income is produced in the sale of the bulleted list of middle return to carry out ranking (for example, giving more weights) to result rather than inciting somebody to action Project for search simple correlation with that may be even more important.Provider may have a large amount of and various information, and (it can be used for pair As a result carry out ranking), for example on its offer resource information, the information on user behavior is (for example, user response is in searching Rope and the frequency for selecting given resource) there is provided business's income information or business partner information.Usually as response user's search As a result resource will be recognized and presented using the part of these information, there is provided business to meet the target of provider.This can be used Information carries out ranking to result, and wherein ranking can provide the order that result is presented to user.

May traditionally be devoted a tremendous amount of time there is provided business attempt to determine its which information for possessing with by reaching The mode of its target is searched related to presentation user's search result.Selected information generally has to be combined, as various work( The input of energy, and weight relative to each other.All these actions are usually directed to provider in each step (for example, identification will make Data, development function and determine function relative weighting) manual intervention.It is used as search or a part for matching algorithm Include risk that is wrong or matching integrity breaks to provide this weighting for inquiring about the search result matched with user.Behaviour Vertical matching weight might have unexpected result.Conversion portion or whole user's inquiries are come by using query rewrite, searched Index is held up and can realize and prevent the optimization of some this emergency risks.In addition, as described above, constraint can be together with query rewrite Use, enable to perform for each user, for each StoreFront, for each geographical position or for other targets Optimization.

Fig. 6 is the example search engine system for showing to carry out ranking for pair search result associated with user's inquiry The block diagram of 600 details.System 600 can include module 604, inquiry front end 606, query node 612 and database 614.Look into Ask node 612 and can also gather comprising order models 610, index of articles 616 and corresponding to inquiring about 602 and as search result The ranking factor 620 of the bulleted list 618 of return is gathered.

Such system can be used available for one or more matching calculation by user inquires about and database items match Method, and can be used for carrying out ranking to user's search result, the result on top is returned into user as search result set Client device.Fig. 6 shows a realization of the component that ranking is carried out to search result.Front end 606 can connect from user Receive inquiry 602.Then, front end 606 can communicate with inquiry factorization module 604, with rewritten query 602 and according to inquiry The 602 generation data factors.Then query profiles can be sent to query node 612, query profiles are substantially by rewriting system Change and add user's inquiry of the added data factor.The example of the data factor can be based on user inquire about in item with And the association between this and the classification of classification tree carrys out the category identifier associated with user's inquiry.

In one exemplary embodiment, system 600 can be the electronics associated with the distribution platform of such as system 700 Commercial search engine.The platform of system 700 can include the StoreFront and the sales platform for trade company of a large amount of trade companies.System 700 can also include auction platform, auction payment system and store front and other E-business services.As all these A part for E-business service, system 700 can include classification tree, and it is used for that by system 700 can sell or auction Product is classified.Such class tree can include top, the wide class in the second level under top of identified category tree, For example, electronic equipment classification, sports equipment classification, class of vehicles or any other such classification.It is each in these classifications The individual constraint being used as in QRIL records as described above.Each two grades of classifications can be included in tree and one or more two The associated one or more three-level classifications of level classification.For example, electronic equipment classification can have three-level classification, the three-level classification Build including TV, computer, smart phone, tablet device and in classification tree its under the electronic equipment classification of the second level Classification as him.In classification tree each bottom level classification or any classification can have with by classification tree classification can be by being Associating key word, metadata or other this type of informations of the product correlation of the sale of system 700.

In addition, triggering and rewriting value may not only be had by rewriteeing, and can there can be classification to rewrite.Classification is rewritten may It can will search for the specific category being limited in classification tree.For example, QRIL records 500, which include classification, rewrites 552.Classification is rewritten can be with It is following rewrite：It will be limited to the specific category of classification tree based on the search that user inquires about, rather than be replaced with rewriting value Trigger mark.For example, QRIL records 500 can include triggering " Brand A television ", wherein it is to be used to delete mark " TV " to rewrite And add the mark adjustment rewriting that classification rewrites " electronic equipment/TV ".Therefore, the use of " Brand A television " is included when receiving When family is inquired about, the rewriting associated with the QRLI for the part that the standardization as search engine rewrites set can be by " brand A TV " is rewritten as such as " brand A：The inquiry of classification=electronic equipment/TV ".Search engine will search for " brand A " items, but only Searched in " TV " classification in classification tree under " electronic equipment ".

When being inquired about by front end receiver to user in 606, user's inquiry can be sent to QFM 604 to be inquired about Rewrite and the generation of the data factor.Query rewrite in QFM 604 can use standardized query as described above to rewrite set.This Outside, data factor generation can recognize the classification associated with user's inquiry.For example, user's phase with submitting user's inquiry 602 The historical data of association can be used for the fuzzy item for distinguishing such as " apple " etc.This classification can be with user's inquiry 602 It is associated, and it is used as the constraint information based on any constraint associated with query rewrite, rewritten as standardization Paris A part for set.In addition, as example described in QRIL records 500, the query rewrite generated is recorded according to QRIL can With including the class label associated with above-mentioned classification tree.In such embodiments, additional queries overwrite type can include mould Classification is pasted to rewrite.Fuzzy category rewrite refer to QFM 604 the data factor generation in keyword, product or classification tree with Other the use that mark in family inquiry or any user inquiry is associated.Therefore, QFM 604 can include be used for One or more modules that family inquiry 602 is classified, its information being then based in module is rewritten for dynamic queries.The letter Breath can include the time-sensitive information associated with trade company sale, holiday sell, the user's history associated with specific trade company Or can in the generation of the data factor any other this type of information that use, its be fuzzy category query rewrite dynamic it is defeated Enter.

Using fuzzy category query rewrite embodiment in, except the every other type available for system rewriting it Outside, system is by including the conflict rule for such rewriting.In one embodiment, for example, fuzzy category inquiry weight Write with lowest priority, and and if only if user's inquiry 602 is just used when other rewritings are not present.In certain embodiments, lead to Conflict between multiple fuzzy queries rewritings occurs for Chang Buhui, because it will be generated by fuzzy rewriting system that fuzzy query, which is rewritten, It is single to rewrite.The single rewriting is generated by obscuring rewriting system based on category analysis or other analysis systems, in other points In analysis system, rewrite based on groups of category associations rather than it is defined from user inquire about mark to rewriting value conversion. On the contrary, fuzzy query is rewritten obscures rewriting module using the inquiry based on the part as the data factor generation in QFM 604 Interior preference information or system are set.

When fuzzy query, which is rewritten, has been allowed to derivation (descendent) or recursive query rewriting, fuzzy query is rewritten can It can cause problem.For example, fuzzy query rewrites the classification that the keyword that user can be inquired about into 602 is rewritten as classification tree.Cause This, is not the mark for searching for user's inquiry, but by for keyword, product or other associated with the classification of classification tree are believed Cease to perform search.This rewriting to classification can also allow for carrying out query rewrite based on the information in classification.As an example, Can be by " brand A " fuzzy query rewriting and the associated categories on " clothes, footwear and accessories/Men's Shoes/motion " Search limitation inquires about " brand A Men's Shoes " to rewrite user.If system also using triggering " brand A shoes ", utilizes rewriting Value " search by brand A " and being directly overwritten to the classification search of classification " clothes, shoes and accessories/Men's Shoes " limitation, then the second conflict Rope will potentially include rewriteeing wider results set than first.As described above, such conflict can be by rewriteeing class Type is prioritized to solve, or by being solved to the rewriting for causing narrower search result set is prioritized.

Received standardization can be used to rewrite set (such as by QRIL processors 140 according to QRIL database of records The standardization of QRIL set of records ends generation in 130 rewrites 142 and gathered) carry out rewritten query.

Query node 612 can apply one or more ranking object modules 610 to query profiles.Such ranking target Model 610 can recognize the match-type for the search result for being qualified as specific inquiry or query profiles.In one example, Object module can be used for selecting search result from database 614.Database 614 can return to the result as inquiry 602 The search index of the bulleted list of return.

The original project data that returns can be included arriving query node 612 by index of articles 616, in query node 612 Place, the non-ranking of list (for example, unordered) of bulleted list 618.The set of the precedence data factor 620 can include for will be by arranging Given bulleted list and all data factors of inquiry 602 that the set of name object module 610 is used.These factors can be transfused to Into ranking object module 610, the ranking result set 622 of user can be then presented to produce.In one example, compared with The bulleted list of high ranking can be significantly more shown (for example, the bulleted list of higher ranked than the bulleted list of relatively low ranking Can be higher than the bulleted list of relatively low ranking in the list of search result of user is presented to.In one example, significantly Ground show higher ranked list can including the use of color (for example, change background or foreground color), animation or other regard Feel decoration (for example, frame, title etc.).

In example 600, search inquiry can be used for the project sold in Online release system or market, but also may be used To consider user's inquiry data resource and carry out ranking and other examples returned to result.The various assemblies of system 600 can be with Performed by software, hardware or its certain combination.In the case of component software, it will be appreciated that perform hard needed for the software Part also there will be.

Fig. 7 be show can with embodiment described herein the block diagram for the network delivery system being used together.Fig. 7 Depict the client-server system 700 that can be used together with various embodiments.For example, can by search engine 160 or System 600 is deployed as a part for system 700.The networked system of exemplary forms with network market or delivery system 702 provide server side function via network 704 (such as internet or wide area network (WAN)) to one or more clients. In some embodiments, trade company can use the registration processes of all operating process described above 305 to issue (system) registration.This The trade company of sample can use system 700 to client device present StoreFront, including as trade company's StoreFront a part provide search Rope engine operation.Then, trade company can provide rewriting, rewriting transcoder generation QRIL records, QRIL notes to transcoder is rewritten Record includes the constraint that the StoreFront of trade company is only applied to when being operated in system 700.Client device then can be to system 700 Search inquiry is submitted, and can be used according to QRIL record generations as the search engine of the part operation of system 700 Standardized query rewrites to generate search result and search result is sent into client device.

For example, Fig. 7 shows (such as browser, such as by Microsoft's public affairs in Redmond, Washington city of web client 706 Take charge of the Internet Explorer browsers of exploitation) and the programming client that is performed on each client machine 710 and 712 708.Client machine 710 and 712 and third-party server 730 can send search inquiry to search engine 723, and search is drawn 723 are held up to operate to provide a user E-business service together with market applied procedure 720.Search engine 723 can use as On in the exemplary embodiment described standardized query rewrite.

Application programming interfaces (API) server 714 and web server 716 are connected to one or more application servers 718, and programming and the web interface for arriving one or more application servers 118 respectively are provided.718 trustship of application server one Or 722 and search engine 723 are applied in multiple markets using 720, payment.Application server 718 is then illustrated as coupled to one Individual or multiple database servers 724, one or more easily accessible one or more databases of database server 724 726。

Market can provide multiple functions of the market and service using 720 to the user of access networked system 702.Pay application 722 can similarly provide a user multiple payment services and function.Paying can allow user to accumulate in the account using 722 Value (such as currency (such as dollar) or proprietary currency (such as " integration ")), is then exchanged with the value of accumulation and is applied by market 720 products being commercially available (such as commodity or service).Although applying 722 using 720 and payment figure 7 illustrates market A part for networked system 702 is all formed, it will be understood that, in an alternative embodiment, paying can form and network using 722 A part for the separated and remote payment services of system 702.

In addition, although system 700 illustrated in fig. 7 uses client-server architecture, but the present invention is certainly not limited to Such a framework, and can equally well be applied to such as distributed or peer-to-peer architecture system.Various markets and payment are applied 720 and 722 and search engine 723 can also be implemented as not necessarily have networked capabilities independent software program, or It is used as the single special equipment via network connection.

Each market of network interface access that networking client 706 is supported by the webserver 716 is using 720 and branch Pay and apply 722.Similarly, the DLL that programming client 708 is provided by API server 714 is accessed to be applied by market The 720 various services and function with payment using 722 offers.For example, programming client 708 can be seller application (for example, by The TurboLister applications of the eBay Inc. exploitations of San Jose), it is used to enable seller with offline The list on networked system 702 is created and managed to mode, and performs batch between programming client 708 and networked system 702 Tupe communicates.

Fig. 7 is also shown in the third-party application 728 performed on third-party server machine 730, the energy of third-party application 128 Enough DLLs provided via API server 714 are programmed access to networked system 702.For example, third-party application 728 The information obtained from networked system 702 can be utilized to support one or more features or the work(on the website being responsible for by third party Energy.For example, third party website can provide by networked system 702 related application support one or more promotion, market or Payment function.

Fig. 8 shows that the diagram of the machine of the exemplary forms with computer system 800 is represented, in the computer system In 800, one group of instruction can be performed so that the machine performs any one or more methods in the method being discussed herein.Example Such as, in certain embodiments, inquiry transcoding device 120, QRIL processors 160 and search engine 160 can be each using calculating The element of machine system 800 realizes above-mentioned special system of computational devices and process.In an alternative embodiment, machine can conduct Autonomous device is operated, or can be connected and (for example be networked) and arrive other machines.In networked deployment, machine can be in service Operated in device-client network environment with the ability of server machine or client machine, or conduct equity (or distribution Formula) peer machines in network environment are operated.Machine can be server computer, client computer, individual calculus Machine (PC), tablet PC, set top box (STB), personal digital assistant (PDA), cell phone, web appliance, network router, exchange Machine or bridger or can (sequentially or otherwise) perform specifies any of the instruction set of action to be carried out by machine Machine.In addition, although individual machine is only shown, but term " machine " should also refer to including individually or jointly execute instruction collection Close (or multiple set) with perform it is described herein any one or more individual method machine arbitrary collection.

Example computer system 800 includes processor 802 (such as CPU (CPU), graphics processing unit (GPU) or both), main storage 804 and static memory 806, they are in communication with each other via bus 808.Computer system 800 may also include video display unit 810 (such as liquid crystal display (LCD) or cathode-ray tube (CRT)).Computer system 800 also include Alphanumeric Entry Device 812 (such as keyboard), cursor control device 814 (such as mouse), disc drive unit 816th, signal generating apparatus 818 (such as loudspeaker) and Network Interface Unit 820.

Disk drive unit 816 is included in storage thereon and embodies one or many in approach described herein or function The machine readable media 822 of individual one or more groups of instruction and data structures (for example, software 824).Held in computer system 800 During row software 824, software 524 can also completely or at least partially be located in main storage 804 and/or positioned at processor In 802, main storage 804 and processor 802 also constitute machine readable media.

Software 824 can also be sent or received by network 826 via Network Interface Unit 820.

Although in the exemplary embodiment, machine readable media 822 is shown as single medium, term " machine readable Jie Matter " is understood to include the single medium for storing one or more instruction sets or multiple media (for example, concentrating or distributed number According to storehouse and/or associated cache and server).Term " machine-readable medium " should also be viewed as including storing, Coding or carrying are executable by a machine and make the instruction set of any one in the method for the machine execution present invention or more The arbitrary medium of conjunction.Term " machine readable media " therefore it should be considered as including but is not limited to：Solid-state memory, optics and magnetic Medium and signal (for example, electric signal or electromagnetic signal).

Therefore, it has been described that use the method and system of the search result rank of machine learning.Although with reference to special Determine example embodiment and describe the present invention, it will be apparent that：The spirit and scope of wider range of invention can not departed from In the case of various modifications and changes are made to these embodiments.Therefore, specification and drawings should be seen as it is illustrative and It is not limited significance.

Although certain embodiments of the present invention has been described in the foregoing specification, and for illustrative purposes Many details are elaborated, it will be clear to those skilled in the art that present subject matter is easily by additional reality The limitation of example is applied, and it is some without departing from the basic principles of the present invention, and details described herein can be considerably Change.

Summary is provided to allow reader quickly to determine the nature of the disclosure of the art and main points.It should be understood that it is not used to solution Release or limit the scope or implication of claim.Therefore, the claims below is merged in embodiment, and each right will Ask and be individually present as single embodiment.

Claims (22)

1. a kind of system, including：

Transcoding device is inquired about, including：

Input module, is configured as receiving the first query rewrite data acquisition system from the first query rewrite source device, wherein, described the One query rewrite data acquisition system includes bound data, meta-tag data and rewrites data, and the bound data at least includes first Trigger value, and the rewriting data at least identify the first query rewrite value associated with first trigger value；

The data parser module of the input module is coupled to, the inquiry data acquisition system of processing first is configured as described to recognize First triggering and the first query rewrite value, and by the inquiry data parsed send to one or more identifier modules with Recognize the first query rewrite type associated with the first query rewrite data acquisition system in multiple queries overwrite type；And

The generation of query rewrite input language QRIL records and formatting module, are configured as from the first query rewrite data set Symphysis is recorded into the first QRIL, wherein, the first QRIL record includes the first trigger value, the first query rewrite value and first yuan The first QRIL record identifications are related to the first query rewrite type by labelled element, the first meta-tag element Connection.

2. system according to claim 1, wherein, the inquiry transcoding device also includes：

Identifier module is constrained, the data parser module is coupled to, and be configured as analysis first query rewrite Data acquisition system is to recognize with first trigger value is applied into the associated one or more constraints of user's inquiry；

Wherein, the QRIL records generation and formatting module are additionally configured to look into described based on the constraint identifier module Ask and rewrite the analysis of data to generate classification binding occurrence, trade company's website binding occurrence as the first QRIL parts recorded Country's value is initiated with inquiry.

3. system according to claim 2, in addition to：

It is coupled to the QRIL database of records of the inquiry transcoding device；

The QRIL processor devices of the QRIL database of records are coupled to, the QRIL processor devices are configured as：Access The QRIL database of records are recorded with retrieving the first QRIL；The first QRIL records are handled to generate the first standardization Rewrite；And, rewrite more new standardized using the described first standardization and rewrite and gather, the standardization to generate renewal, which is rewritten, to collect Close；And

The search engine system of the QRIL processors is coupled to, wherein, the search engine system is configured as：Receive described The standardization of renewal rewrites set；Rewrite to rewrite the first search from the first client device using the described first standardization Engine owners inquiry is inquired about with the user for creating rewriting；The first search result set is generated for user's inquiry of the rewriting Close；And, send first search result set to first client device from the search engine.

4. a kind of computer implemented method, including：

The first query rewrite data acquisition system is received from the first query rewrite source device at inquiry transcoding device, wherein, described the One query rewrite data acquisition system includes bound data, meta-tag data and rewrites data, and the bound data at least includes first Trigger value, and the rewriting data at least identify the first query rewrite value associated with first trigger value；

Processing first inquires about data acquisition system to recognize first triggering and the first query rewrite value；

Analyze it is described first inquiry data acquisition system with recognize in multiple queries overwrite type with the first query rewrite data set Close the first associated query rewrite type；

The first query rewrite input language QRIL records are generated from the first query rewrite data acquisition system, wherein, described first QRIL records include first trigger value, the first query rewrite value and the first meta-tag element, first meta-tag The first QRIL record identifications are associated with the first query rewrite type by element；And

The first QRIL records are stored in the QRIL database of records recorded with multiple QRIL.

5. method according to claim 4, in addition to：

The QRIL database of records as described in QRIL processor die block access, are recorded with retrieving the first QRL；

The first QRIL records are handled using the QRIL processor modules, to generate the first standardization rewriting；And

Rewritten using the described first standardization and carry out the rewriting set of more new standardized, the standardization to generate renewal, which is rewritten, gathers.

6. method according to claim 5, in addition to：

The standardization that the renewal is received at the inquiry factorization module of search engine rewrites set；

In the search engine inquiry of the first search engine user is received from the first client device；

Rewritten using the described first standardization and the inquiry factorization module rewrites first search engine user and inquired about；

The first search result set that generation is inquired about for first search engine user is rewritten using the described first standardization； And

First search result set inquired about for first search engine user is sent to from the search engine First client device.

7. method according to claim 4, wherein, first triggering includes first group of mark, and first group of note Each mark in number includes the character for forming a word.

8. method according to claim 4, wherein, the first query rewrite value includes more than second mark, and described the Each mark in more than two mark includes the character for forming a word, and more than first mark is different from more than described second Each mark in individual mark, more than second mark is the synonymous of at least one mark in more than first mark Word.

9. method according to claim 7, wherein, the multiple query rewrite type includes：

Phrase overwrite type；

Mark refines type；

It is directly overwritten type；And

Full query rewrite type.

10. method according to claim 7, wherein, the first query rewrite type is to be directly overwritten type, described One triggering includes the first set of letters, and it is that search engine connects that the first query rewrite value, which is included in first set of letters, The second set of letters that first set of letters is used, institute are replaced during searching for during a part for the user's inquiry received State the second set of letters different from first set of letters.

11. method according to claim 7, wherein, the first query rewrite type is that token refines type, and described First query rewrite value mark from client device receives described the when the part that inquired about as user in search engine Will be in the first token removed during searching for from the multiple token in search engine during one triggering.

12. method according to claim 7, wherein, the first query rewrite value is also identified when the search engine is made For user inquire about a part from client device receive it is described first triggering when will be in the search phase in the search engine Between be added to the second token of the multiple token.

13. method according to claim 4, wherein, the first query rewrite type is phrase overwrite type, described to touch Give out a contract for a project and include multiple tokens according to defined order, and the first query rewrite value include when the search engine as with A part for family inquiry will replace described many when receiving multiple tokens according to defined order from client device One or more tokens that individual token is used.

14. method according to claim 4, wherein, the first query rewrite type is full query rewrite type, and institute The first query rewrite value mark is stated when user's inquiry that search engine is received from client device is accurately matched with the triggering When the token value set that will be used by the search engine as dedicated search token.

15. method according to claim 4, wherein, the country of bound data mark first；And

Wherein, the first QRIL records include country's constraint element, and country's constraint element assignment is only when in search engine When place receives the triggering as a part for the inquiry for being identified as initiating in first country by the search engine Just the first query rewrite value can be applied to the triggering.

16. method according to claim 4,

Wherein, the triggering includes the token set of letters of first language；

Wherein, the first query rewrite value includes the token set of letters of the second language associated with the described first country； And

Wherein, the first query rewrite value includes the token for the trade company that mark is located in first country.

17. method according to claim 4,

Wherein, the bound data identifies the first merchant website；And

Wherein, the first QRIL records include website restrictions element, and the website restrictions element assignment is only when in search engine Place can just be looked into when receiving the triggering as a part for the inquiry associated with first merchant website by described first Ask rewriting value and be applied to the triggering.

18. method according to claim 17, in addition to：

First Merchants register associated with first merchant website is received at registration equipment；

By the registration device authentication mandate associated with first merchant website；

Authorized in response to described, the first query rewrite source device is associated with first merchant website；

In response to receiving the first query rewrite data acquisition system from the first query rewrite source device, automatically generate for first The website restrictions element of QRIL records；

After the first QRIL records are generated, the second query rewrite data set is received from the first query rewrite source device Close；And

In response to determining to receive the second query rewrite data acquisition system from the first query rewrite source device, automatically give birth to Into the website restrictions element of the 2nd QRIL for being generated according to the second query rewrite data acquisition system.

19. a kind of computer-readable medium including computer-readable instruction, the computer-readable instruction makes when executed Obtain one or more processors：

The first query rewrite data acquisition system is received from the first query rewrite source device at inquiry transcoding device, wherein, described the One query rewrite data acquisition system includes bound data, meta-tag data and rewrites data, and the bound data at least includes first Trigger value, and the rewriting data at least identify the first query rewrite value associated with first trigger value；

Processing first inquires about data acquisition system to recognize first triggering and the first query rewrite value；

Analyze it is described first inquiry data acquisition system with recognize in multiple queries overwrite type with the first query rewrite data acquisition system phase First query rewrite type of association；

The first query rewrite input language QRIL records are generated from the first query rewrite data acquisition system, wherein, described first QRIL records include first trigger value, the first query rewrite value and the first meta-tag element, first meta-tag The first QRIL record identifications are associated with the first query rewrite type by element；And

The first QRIL records are stored in the QRIL database of records recorded with multiple QRIL.

20. computer-readable medium according to claim 19, wherein, the instruction also causes one or more of places Manage device：

Received at the inquiry transcoding device from Online e-business search engine with that can be obtained via network business system The classification tree that the product arrived is associated, the network business system is associated with the Online e-business search engine；

The first query rewrite data acquisition system and the classification tree are analyzed, to recognize the related to the triggering of the classification tree The first category of connection；

Generation categorization metadata mark is used as a part for the first QRIL records, wherein, the categorization metadata mark Identify the first category of the classification tree.

21. a kind of system, including for the dress for each step for performing the method according to any one of claim 4 to 18 Put.

22. a kind of computer-readable medium of load capacity calculation machine readable instruction, the computer-readable instruction makes when executed One or more processors perform the method according to any one of claim 4 to 18.